across-the-board phenomena › dvries › pdf › 2015-atb-companion.pdf · atb movement rather...
TRANSCRIPT
1
Across-the-Board Phenomena Mark de Vries
University of Groningen
Abstract
Across-the-Board phenomena in linguistics, ATB for short, can generally be defined as one-to-many
correspondences. Standard examples involve wh-extraction out of two or more coordinated clauses or
phrases at once, as in Which teacher did John like _ and Sue hate _? Such configurations violate an
otherwise valid bi-uniqueness principle: an antecedent licenses one and only one dependent element;
furthermore, they are at variance with the alleged island status of coordinate structures. As such, ATB
phenomena constitute an intriguing syntactic puzzle. There are diverse views on the nature and explanation
of ATB, as well as on the question of which construction types can be subsumed under it. Coordination is
clearly an important factor, since ATB does not normally involve hierarchically different positions, with the
possible exception of parasitic gaps and a few other construction types. It has also been claimed that ATB
phenomena extend to certain cases of ellipsis, including right node raising. Apart from this, it is evident that
various types of features, including case, can be spread over multiple conjuncts, and that binding by some
antecedent outscoping the entire coordination phrase may concern pronominal variables in more than one
conjunct at once. This raises the question of whether some kind of multidimensional representation is
required for coordinate structures. Related topics to be discussed from the current perspective are forward
versus backward multiple dependencies, and split antecedents.
Keywords: coordination, multiple dependencies, parallelism, symmetry, asymmetry, wh-movement, ellipsis,
multidominance, scope, binding, variable, split antecedent, parasitic gap
1 Introduction
Ross (1967: §4.2.4) first described Across-the-Board phenomena (ATB) as a class of rules that
move a constituent out of all the conjuncts of a coordinate structure at once. To these parallel
movement rules, the Coordinate Structure Constraint (CSC) does not apply. Ross originally focused
on backward conjunction reduction and relative clause formation. Williams (1978), in an important
elaboration, adds wh-movement in embedded questions to the picture, as well as instances of
forward ellipsis, namely conjunction reduction and comparative deletion. Some of the original
examples will be cited below. Nowadays, standard examples can be considered wh-movement
constructions as in (1):
(1) Which book does Peter like and Susan hate?
Here, the wh-object is extracted from two conjoined clausal projections at once. This is remarkable,
since the CSC says that “In a coordinate structure, no conjunct may be moved, nor may any element
contained in a conjunct be moved out of that conjunct” (Ross 1967: §4.2.1). Although the precise
formulations and even the existence of both the CSC and ATB are now contested (for discussion,
see Gazdar 1981, Steedman 1985, Wilder 1994, Zhang 2004/2010, and Chaves 2012, among
others), it is descriptively clear that coordinate structures can be subject to parallelism requirements
(setting complications aside for a moment), and therefore potentially block certain syntactic
dependencies unless these apply to every conjunct simultaneously. This is illustrated in (2), where
(2a) (=1) represents the ATB pattern:
(2) a. Which book does [Peter like _ ] and [Susan hate _ ] ?
b. * Which book does [Peter like _ ] and [Susan hate Moby Dick] ?
c. * Which book does [Peter like Moby Dick] and [Susan hate _ ] ?
2
In (2a), the fronted wh-phrase which book can be related to the object position of both like and hate,
that is, to a variable in both conjuncts between brackets (indicated by underscores). In (2b) and (2c),
extraction takes place out of only one conjunct, and this is ruled out in semantically symmetrical
configurations. Notice that it is not the case that conjuncts must be internally parallel to each other
in every respect: although clause types are normally synchronized, coordinated clauses sometimes
differ with respect to voice, tense, and aspect, or the positioning of arguments, adjuncts, and
particles. However, potential relationships with an outside element must be similar – with certain
exceptions, to be discussed below.
Parallelism requirements aside, ATB clearly shows that coordinate structures themselves
(conjunction phrases in X-bar terms) are not inherent barriers for movement or grammatical
dependencies more generally. This is corroborated by the fact that ATB is blocked whenever the
corresponding movement in a singular structure would be blocked; see (3), for instance:
(3) a. * Which book do you know [a man that likes _ ] ?
b. * Which book do you know [a man that likes _ ] and [a woman that hates _ ] ?
Extraction from a relative clause in a complex noun phrase is normally banned, and this is no
different in ATB configurations. Therefore, ATB is not simply an exception to constraints on
movement. Rather, it represents an interesting consequence of the grammar’s capacity to duplicate
relationships by means of coordination.
This change of perspective is also useful when the more controversial case of backward
conjunction reduction is studied, illustrated in (4). Here, the object again relates to two verbs at
once, but now the dependency is backwards.
(4) Peter likes, and Susan hates, this book.
In Ross (1967), Postal (1974), and later work, such constructions are conceived of as rightward
ATB movement rather than backward ellipsis. This immediately shows that it depends on
definitions and analyses what counts as a proper ATB phenomenon.
Generally, Across-the-Board phenomena can be defined as one-to-many correspondences:
one operator or antecedent licenses two or more variables or otherwise dependent elements. It is
interesting that the grammar seems to allow this possibility, since it is in contradiction with a simple
principle of bijection or bi-uniqueness that might fit the theoretical ideal.1
This chapter is organized as follows. Section 2 discusses regular ATB movement
constructions and analyses thereof. Section 3 highlights other (potential) instances of ATB
phenomena, including ellipsis, parallel feature checking, and split antecedents. Section 4 is the
conclusion. For reasons of space, the majority of the examples are in English, but similar data can
be produced in many other languages as well.
1 See Koopman & Sportiche (1982) for a specific implementation concerning operator-variable relations, or
Chomsky’s (1981) Theta Criterion, which is also bidirectional. But the principle could be formulated more generally for
any kind of dependency; see also Wilder (1994) and especially Koster (1999, 2003). Although it is generally accepted
that a dependent element must have exactly one antecedent, the inverse is not. In Neeleman & Van de Koot’s (2002)
Bare Phrase Structural derivation of Koster’s (1987) original ‘Configurational Matrix’, for instance, it is taken for
granted that one-to-many correspondences are possible – with reference to ATB phenomena, unsurprisingly. Koster
himself in his more recent work simply sets aside such facts as exceptions that may involve two links rather than one
complex one, while stressing the importance of bi-uniqueness in other respects. See section 3 for more discussion.
3
2 Leftward ATB movement
2.1 A concise empirical survey
ATB wh-movement can be found in a variety of wh-constructions. The examples in (5) show this
for direct and embedded wh-questions, and relative clauses:2
(5) a. Which book did you say that Peter likes _ and Susan hates _?
b. I wonder which book Peter likes _ and Susan hates _.
c. I know a book which Peter likes _ and Susan hates _.
Williams’ (1978: 31) original example fragment was who John saw and Bill hit.
Wh-extraction out of coordinated CP-clauses is a possibility, but it also works if the
coordination is at a lower level, say of IP, VP, or even DP constituents:
(6) a. Who(m) did you say [(that) Peter likes _ ] and [(that) Susan hates _ ]?
b. Which teacher does [Peter like _ ] and [Susan hate _ ] ?
c. What did Peter [buy _ last week] and [throw away _ yesterday] ?
d. Which famous scientist did Peter read [a book by _ ] and [a newspaper article about _ ] ?
In some cases it is not immediately clear if the sentence involves ATB from coordinated phrases
(7a) or simply coordination at the word level (7b).3
(7) a. Which book did John [buy _ ] and [read _ ]?
b. Which book did John [buy and read] _ ?
In what follows, potential ambiguity of this kind is avoided.
The moved phrase does not need to be an object or even a noun phrase: in principle, it can
have any syntactic function and any categorial status. This is illustrated in (8a-c):
(8) a. Which man did you say _ bought a new car and _ sold his boat? (subject, nominal)
b. How fast did John speak _ and Mary write _ during the meeting? (adjunct, adjectival)
c. In whose car did John fall asleep _ and Mary vomit _ ? (adjunct, prepositional)
Potential symmetry requirements are discussed in the next section.
ATB extraction is also allowed with other kinds of A´-movement, including topicalization
(9a-b). This implies that the ATB-moved constituent does not need to be a wh-phrase. Example (9b)
is presented in Dutch, a verb second language, where it is clearer than in English that the adverbial
phrase is actually moved.
(9) a. This man, Peter wants to meet _ but Susan prefers to avoid _.
2 Salzmann (2012a) also provides examples from Zurich German in which one or both of the relative gaps can be
replaced by a resumptive pronoun (in that case, there is no relative pronoun, but an invariant complementizer wo).
On a historical note, Ross (1967) discusses ATB relative clause and question formation, but not wh-movement in the
sense of (5). There appears to be some confusion about this in the literature. His example (i), for instance, crucially
contains two wh-elements; what is construed ATB, apparently, is the relative head NP, which has a copy representative
in each embedded conjunct according to Ross’ description. This is depicted in (ii), from which the surface
representation in (i) is to be derived:
(i) Students who fail the final exam or who do not do the reading will be executed. (Ross 1967: §4.2.4.1)
(ii) [NP [NP Students] [S or [students fail the final exam] [students do not do the reading]]] will be executed. 3 However, the option of head coordination has been contested on independent grounds (Kayne 1994: 59-63).
4
b. Op zondagen wil Peter _ feestvieren en Susan _ studeren.
on Sundays wants Peter party and Susan study
‘On Sundays, Peter wants to party, and Susan wants to study.’
Also note that ATB relativization may involve a null relativizer in English (10a), or a d-relative
pronoun as in the Dutch equivalent (10b).4
(10) a. These are the books OP (that) Peter wrote _ and Susan admired _.
b. Dit zijn de boeken die Peter _ heeft geschreven en Susan _ bewonderd.
these are the books DEM:REL Peter has written and Susan admired
But ATB is not limited to A´-movement. The following examples show A-movement in raising and
passive contexts, respectively:
(11) a. Peter seems to _ like plays and to _ go to the theater quite often.
b. This book is written _ by Peter and illustrated _ by Susan.
ATB head movement is also attested. In (12a/b), the finite aspectual, temporal, or modal verb is
extracted from both conjuncts:
(12) a. Never has Peter _ eaten pork or Mike _ drunk alcohol.
b. Never will/may Peter _ eat pork or Mike _ drink alcohol.
ATB scrambling of an object across adverbs can be illustrated in Dutch:
(13) Susan heeft dit boek gisteren _ gekocht en vandaag _ gelezen.
Susan has this book yesterday bought and today read
‘Susan bought this book yesterday and read it today.’
Similarly, ATB clitic extraction is possible in Romance and Slavic languages. The following
example is from European Portuguese, cited from Matos (2000:233):
(14) Todos o viram _ na aula e cumprimentaram _ delicadamente.
all himCL saw in.the classroom and greeted politely
‘They all saw him in the classroom and greeted him politely.’
In English, so-called tough movement can take place in ATB fashion:5
(15) Love is hard to find _ and easy to lose _.
Thus, it is clear that ATB phenomena are not limited to wh-movement, but can involve any possible
kind of extraction.
4 Relative d-words in various Germanic languages are not complementizers but pronouns, unlike English that. This
is evident from agreement paradigms, pied piping, etc. What is confusing is the sometimes homophonous demonstrative
form of the complementizer. See De Vries (2002:155ff) for discussion and further references. 5 Chomsky (1981) analyzes this complex construction type in terms of operator movement. For more recent
discussion, see Hicks (2009) and Zwart (2012), among others.
5
2.2 Symmetry and asymmetry
2.2.1 Conjuncts and parts of conjuncts
The CSC consists of two parts that turn out to have a different status (see Grosu 1973 and others
since). These may be referred to as the Conjunct Constraint (“no conjunct may be moved”)6 and the
Element Constraint (“no element of a conjunct may be moved out of the coordinate structure”). The
ATB ‘exception’ naturally applies to the latter only; if full conjuncts were ATB-moved, the result is
both nonsensical and nearly impossible to parse. Compare (16a) to (16b):
(16) a. * Whoi did you meet _i and _i ?
b. * Did you meet Johni and Johni?
The wh-operator in (16a) needs to license both variables. But this leads to vacuously coordinating a
phrase to itself, as in (16b) – or a gap to a gap from a surface perspective; see also Bouma, Malouf
& Sag (2001) for comments.
2.2.2. Syntactic functions and theta roles
Even when coordination is underlyingly symmetric, not every type of ATB extraction is allowed. At
first sight, the dependency must involve the same syntactic function in each conjunct; see (17), for
instance, where who corresponds to the subject in the first conjunct and the object in the second, or
vice versa.
(17) a. * Who did you say [Peter likes _ ] and [ _ hates Susan] ?
b. * I met a woman who [ _ hates Susan] and [Peter likes _ ].
In languages with morphological case marking, it is evident why this is: the moved phrase would
have to have express two conflicting forms; see also Dyła (1984) and Franks (1993) for discussion.
But even if certain case differences are not signified, as in the English examples here, the restriction
remains. This is remarkable because syncretism generally precludes matching effects (see the
Companion to Syntax chapter on Free relatives, for instance). In addition, Williams (1978: 34)
notes that subject/object ATB is fine if the subject is embedded; see the contrast in (18):
(18) a. * I know a man who [Bill saw _ ] and [ _ likes Mary].
b. I know the man who [John likes _ ] and [we hope _ will win].
6 Nonetheless, right-extraposition of conjuncts is possible. Most likely, however, this involves (sub-)clausal
stripping (i.e., ellipsis), and hence it provides no counterexample to the Conjunct Constraint (pace Johannessen 1998:
215-217). Arguments for this view include the distributivity effects illustrated in (i)-(iii). These data follow
straightforwardly from an ellipsis account, in which the a.-cases are derived from c. rather than b.:
(i) a. John _ seem*(s) nice, and Mary.
b. John and Mary seem(*s) nice.
c. John seem*(s) nice and Mary seem*(s) nice.
(ii) a. * John _ met, and Mary.
b. John and Mary met.
c. * John met and Mary met.
(iii) a. * John _ talked to each other, and Mary.
b. John and Mary talked to each other.
c. * John talked to each other and Mary talked to each other.
See Merchant (2003) and Chaves (2012:468) for a similar assessment and pertinent remarks. See De Vries (2002:233ff)
and subsequent work for a concrete (and general) proposal of right-extraposition in terms of ellipsis.
6
According to De Vries (1992) cases like (18a) are unacceptable simply due to parsing problems,
and not due to restrictions of the grammar, as proposed in Williams (1978) or Pesetsky (1982).
Concretely, the incorrect parse Bill [saw and likes] Mary is possible within the embedded clause,
but clashes with the context. This view is supported by Chaves (2012:482), who notes that (19) is
acceptable despite the fact that the subject is not embedded.
(19) There were some parts that [[I enjoyed _ ] and [ _ were very suspenseful]].
Here, no garden path arises because I [enjoyed and were very suspenseful] is impossible.
The following examples show that direct objects can be combined with prepositional objects
in ATB constructions:
(20) a. What did you say [Peter stared at _ ] and [Susan desired _ ] ?
b. Which tree did you say [Peter hugged _ ] and [Susan talked to _ ] ?
c. Which criminal did you say [the citizens feared _ ] and [the police kept away from _ ] ?
ATB movements involving an indirect object in double object constructions in combination with
other functions can be illustrated in Dutch (generally, non-prepositional indirect objects cannot be
wh-moved in English, which bleeds more complex examples):
(21) a. Wie heb jij _ geld geleend en _ kleren gegeven? (IO+IO)
who have you money lent and clothes given
‘Who did you lend money [to] and give clothes [to]?’
b. Welke man heb jij uiteindelijk _ het jawoord gegeven, maar (IO+DO)
which man have you finally the consent given but
heeft Anna al die jaren _ begeerd?
has Anna all those years desired
‘Which man did you say “I will” to in the end, but did Anna desire all those years?’
c. Welke astronaut is een vlucht toegezegd, maar _ heeft nog geen ruimtepak? (IO+S)
which astronaut is a flight promised but has yet no space.suit
‘Which astronaut:IO was a flight:SU promised [to], but does not have a space suit, yet?’
d. ? Wie zei je dat er _ niet kan lezen, maar jullie toch _ een boek (S+IO)
who said you that there not read can but you yet a book
hebben gegeven?
have given
‘Who did you say cannot read, but you:PL gave a book [to] anyway?’
Though often symmetry-violating examples are awkward to some degree (and speakers may judge
differently), the fact that acceptable examples can be construed implies that limitations are likely
due to extraneous factors such as parsing problems and pragmatics, rather than deep grammatical
principles.
In concert with these findings, it is to be noted that different theta roles may be mingled. In
(22), who represents not only an experiencer, but also an agent, theme, and beneficiary. It is the
subject of various kinds of predicates at once: transitive, unergative, unaccusative, and raising.7
(22) Who did you say [ _ saw a crocodile], [ _ ran for an hour], [ _ arrived late], [ _ seemed to cry],
and [then _ received the consolation prize] ?
7 Similar issues need to be investigated with respect to Right Node Raising, but this cannot be entertained in detail
here.
7
This state of affairs yields an important question: how is it possible that apparently one phrase is
assigned two or more potentially different thematic roles and syntactic functions? Answering this
question is a key to understanding ATB.
2.2.3 Embedding and islands
As was mentioned in the introduction, ATB wh-extraction is island-sensitive (see (3)). The
movement involved is also unbounded; this is shown in (23):
(23) Which book do you think [that Peter said [he bought _ ]] and [that Susan claimed [she actually
read _ ]] ?
To prevent garden paths, the intended structure can be facilitated by intonational grouping and, to
the extent possible, the strategic placement of optional complementizers, as is done in the English
examples here.
Conjuncts do not need to be equally complex, but may neither constitute an island. In (24c/d),
one of the conjuncts contains a complex noun phrase.
(24) a. Which book do you think [Peter bought _ ] and [Susan claimed [she actually read _ ]] ?
b. Which book do you think [that Peter said [he bought _ ]] and [that Susan actually read _
] ?
c. * Which book do you think [Peter bought _ ] and [Susan knows [a man who actually
read _ ]] ?
d. * Which book do you think [that Peter knows [a man who bought _ ]] and [that Susan
actually read _ ] ?
Zhang (2010:226) presents the following pair (annotation added here) that appears to show an
asymmetry between the first and second conjunct. However, other speakers consulted found (25b)
unacceptable, which matches the situation in (24).
(25) a. * Who did [Bill lose business [because he hired _ ]] and [Mary praise _ a lot] ?
b. %
Who did [Bill praise _ a lot] and [Mary lose business [because she hired _ ]] ?
All in all, there does not seem to be convincing evidence for locality asymmetries, and to the extent
there are any, they are speaker-dependent.
2.2.4 Referential identity?
Another symmetry issue is the question of referential identity of an ATB phrase itself. It seems that
the most salient reading in (26) is one where the variable bound by the wh-phrase is identical in
both conjuncts, and receives a strict interpretation.
(26) Which book/what/who does Peter like _ and Susan hate _?
Nevertheless, speakers of both English and Dutch consulted report an available respectively answer
to questions like this, depending of course on a plausible discourse. In (27a), such a sloppy reading
is even forced (a song cannot be sung before it has been composed, so different songs must be
involved), and the sentence is perfectly acceptable, e.g., in the context of two available lists of
8
songs, one pertaining to compositions and one to performances.8 Another example is (27b), where it
seems plausible that the number of matches involved is different for Peter and Susan.
(27) a. I wonder which song Peter composed _ today and Susan sang _ yesterday.
b. How many matches did Peter play _ and Susan win _?
Thus, contrary to suggestions in the literature (see especially Zhang 2010), there is no ‘identity
requirement’ in ATB.9
Referential identity can also be tested in topicalization constructions. The uniqueness property
of definites automatically leads to a strict identity reading (28a). A specific indefinite (28b) usually
receives a strict interpretation, but a sloppy one is not necessarily excluded, as before. The plural
indefinites in (28c/d) receive a generic interpretation; in that case there is identity of the kind
function, but not of individual entities. Finally, the modal context in (28e) facilitates a nonspecific
reading for the indefinites; one could say that this automatically leads to a sloppy interpretation.
(28) a. This/my/the car, Peter admired _ and Susan bought _.
b. A red car, Peter admired _ and Susan bought _.
c. Such cars, the poor man desires _ and the rich man can afford _.
d. Cats, Peter likes _ and Susan hates _.
e. A red car / red cars, Peter would like to have _ and Susan would hate to have _.
Furthermore, when a complex topicalized phrase contains a pronoun, a sloppy reading of the
embedded variable becomes readily available:
(29) a. Pictures of himselfi resp. j, Hollandei hates _ but Putinj loves _.
b. Hisi resp. j own mother, Peteri likes _ and Johnj hates _.
c. Calling hisi resp. j mother, the winneri of the tennis match attempted _ but the loserj
avoided _.
This is similar in wh-constructions: (30a/b) are based on Haïk (1985: 286) and Ha (2008:254),
respectively.10
(30) a. I wonder which picture of himselfi resp. j Peteri likes _ and Johnj hates _.
8 This apparently contradicts Heycock & Zamparelli (2000:351), where a similar example is tagged #, but according
to their description the intended reading is merely unlikely, not necessarily impossible. 9 In English, an emergent, non-reconstructed plural can sometimes be used to force a respectively reading; see
Gawron & Kehler (2004), who expand on Munn (1999:421):
(i) Q In what cities did Mary vacation _ and Bill decide to live _ ?
A Mary vacationed in Paris and Bill decided to live in Toronto.
Munn argued that such questions involve a complex functional reading; consequently, a supposed ATB identity
requirement could target this function rather than its argument index (note that a similar reasoning may apply to kind
readings as in e.g. (28c/d) below). According to Zhang (2010), such instances of non-identity in ATB configurations
must be theoretically set apart from the regular ones, requiring a different analysis, in which the preposed phrase is
generated outside of the coordination; see also Woolford (1987) and Chaves (2009) for some related discussion
concerning ATB adjunct extractions. However, the widespread availability of sloppy readings presented in the main text
– also in regular arguments (i.e., non-adjuncts) – casts serious doubt on this approach. 10
Both Haïk (1985) and Munn (1992:10) report a curious asymmetric strict reading for some English speakers; see
(i). This requires himself to be used logophorically in the second conjunct.
(i) Which picture of %
himselfi/*herselfj did Johni buy _ and Maryj paint _ ?
Note that with a proper local anaphor such as Dutch zichzelf (which is underspecified for gender), this is impossible,
and only a sloppy reading can be obtained: Welk schilderij van zichzelfi resp. j heeft Johni gekocht _ en Maryj geschilderd
_?
9
b. Q Whose car does Johni like _ but Billj hate _ ?
A Hisi resp. j own car.
In conclusion, wherever non-identity readings are available on independent grounds, they can be
obtained in ATB constructions.
2.2.5 Multiple ATB
So far, most illustrations involved two conjuncts, but there is no principled boundary to the number
of dependencies (see also (22) above). Example (31a) involves a quadruple instantiation of ATB
wh-movement; triple ATB in a non-restrictive relative clause (31b) is taken over from Pesetsky
(1982: 444).
(31) a. Who does Peter like _ , Susan admire _ , Mike hate _ , and John detest _ ?
b. John, who I bought a picture of _ , a story about _ , and a book by _
Every conjunct necessarily participates in the particular dependency, and interruptions of
parallelism as in (32) are generally disallowed. Compare (32) to (31a):
(32) * Who does Peter like _ , Susan admire the president, Mike hate _ , and John detest _ ?
Somewhat similarly, if more than one constituent is ATB-moved, the double dependency must be
present in each conjunct; see the following examples, slightly adapted from Pesetsky (1982: 445):
(33) a book OPi that I know whoj to...
a. ... [[talk to _j about _i] and [persuade _j to buy _i]].
b. * ... [[talk to Bill about _i] and [persuade _j to buy _i]].
c. * ... [[talk to _j about Mary] and [persuade _j to buy _i]].
However, parallelism requirements are weakened in semantically asymmetric coordinations, as is
documented in Lakoff (1986), for instance:
(34) How many courses can you [take _ for credit], [still stay sane], and [get all A’s in _ ] ?
It is debated whether asymmetric coordinations are true coordinate structures at all levels of
representation, and also if extraction is real in these cases: according to Postal (1998), a silent
resumptive pronoun may be involved (however, see Levine 2001 for a critical reply). Such issues
are beyond the scope of this chapter.11
2.2.6 ATB in multiple wh-fronting languages
Some languages allow fronting of more than one wh-phrase in multiple questions, unlike the
situation in English. This provides an interesting testing ground for ATB behavior. As Citko
(2011:57) illustrates for Polish, double spell-out of an ATB-extracted wh-phrase is generally
disallowed (35a), whether i=j or i≠j, but what can be done is ATB extraction of two different wh-
phrases from both conjuncts (35b):
11
Examples of asymmetric extraction were already discussed in Ross (1967); see also Schmerling (1972).
Goldsmith (1985), Lakoff (1986), Culicover & Jackendoff (1997), and Kehler (2002) concluded that the CSC cannot be
a syntactic constraint. This view is opposed by Postal (1998) and De Vos (2005). See Zhang (2010) and Chaves (2012)
for a recent overview, as well as the separate Companion to Syntax chapter on Asymmetric Coordination. See also
section 2.2.7 and the related footnotes 12 and 13.
10
(35) a. * Coi coj [Jan zgubił ti ] a [Piotr znalazł tj] ?
what what Jan lost and Piotr found
‘What did Jan lose and Piotr find?’
b. Coj komui [Jan kupił tj ti ] a [Piotr wysłał tj ti ] ?
what whom Jan bought and Piotr sent
‘What did Jan buy for whom and Piotr send to whom?’
Extraction of two distinct wh-objects from the same position in separate conjuncts is unacceptable,
as noted in Kasai (2004:169) for Serbo-Croatian and other languages; see (36a). What is attested,
however, is regular subject question formation in combination with ATB wh-extraction of an
embedded object from multiple conjuncts. The Serbo-Croatian example in (36b) is taken from
Bošković & Franks (2000:111).
(36) a. * Kogai štaj on [vidi ti ] i [jede tj] ?
whom what he sees and eats
‘[*] Whom what does he see and eat?’
b. Koi štaj ti tvrdi [da Jovan kupuje tj ] i [da Peter prodaje tj ] ?
who what asserts that John buys and that Peter sells
‘[*] Who asserts that John buys what and that Peter sells what?’
In short, ATB in multiple fronting languages may result in somewhat complex patterns, but in fact
behaves as one would expect.
2.2.7 Covert ATB?
Interestingly, multiple questions result in partial wh-in-situ in English and many other languages
(37a), but doubling of this situation across conjuncts is excluded (37b). This contrasts notably with
(36b).
(37) a. Who said that John bought what?
b. * Who said [that John bought what] and [that Peter sold what] ?
According to Bošković & Franks (2000), such data suggest that ATB movement cannot take place
covertly at LF. This appears to be confirmed by their example (38b), where only the low scope
reading for the universal quantifier is available, whereas it can normally obtain a high scope reading
(38a) – a fact often explained in terms of covert Quantifier Raising:
(38) a. Some boy hugged every girl. ( > or > )
b. Some boy [hugged every girl] and [kissed every girl]. (only > )
However, a wide scope reading for every is available in Right Node Raising contexts such as (39);
see also Sabbagh (2007):
(39) A (different) reviewer read through _ and passed judgment on, every submission.
Furthermore, RNR repairs the type of problem noted in (37b):
(40) a. Who said that John bought (_/*what) and Peter sold what?
b. What explains (_/*what) or provides an excuse for what?
c. Who received bribes from (_/*whom) or was threatened by whom?
11
It seems, then, that (covert) ATB hinges on sharing of the phrase to be moved. In (37b) and (38b)
this is not the case.
According to Mayr & Schmitt in the chapter on Asymmetric Coordination, covert ATB
movement is necessary in so-called SLF-coordination in V2-languages (‘subject lacking in finite
clauses’; see Höhle 1983).12
The relevant German example is translated into Dutch here:
(41) Helaas heeft geen van mijn kennissen een hond of _ kan met katten
unfortunately has none of my acquaintances a dog or can with cats
omgaan.
deal
‘Unfortunately, no acquaintance of mine has a dog and no acquaintance of mine knows how
to deal with cats.’
If they are right that coordination is at the C´-level, the overt subject geen van mijn kennissen ‘none
of my acquaintances’ must move out of the first conjunct in order to outscope the coordinator or
(note that this is obligatory, which contrasts with the situation in (38a), where the reverse scope
reading is optional). However, if coordination were below the subject – i.e., at the I´-level, cf. Zwart
1991 in particular –, the problem of scope does not arise in the first place, and there is no need for
covert ATB in these examples.13
2.2.8 Parasitic gaps vs. successive cyclic movement
ATB-extractions in coordination contexts may be compared to parasitic gap constructions, which
also involve a single visible wh-operator and two dependencies (Taraldsen 1981, Engdahl 1983). A
typical example is (42):
(42) Which book did Peter burn _ [without reading _ ] ?
Here, the gap inside the PP adjunct – an island for extraction – is said to be parasitic on the object
trace. Without regular wh-movement, there can be no gap inside the PP; see (43a), where the
pronoun it is obligatory. By contrast, movement of the object of the matrix verb is not dependent on
a gap inside the adjunct; see (43b), where it is optional:
12
A perhaps related piece of data concerns the phenomenon of ‘clitic omission’ in Czech. In (i), cited from Dotlačil
(2008), the clitic object ho ‘him’ is interpreted in both clauses, but visible only in the first:
(i) Zavolal jsem ho a představil _ známým.
called AUX:1SG him and introduced friends
‘I called him and introduced him to friends.’
According to Dotlačil, the shared clitic is covertly ATB-extracted, and for some reason spelled out in its base position
inside the first conjunct only.
This differs from other types of object gaps. For instance, in Dutch, (ii) is possible for some speakers. This
requires an intonational pattern that is similar to the one found in gapping constructions. Crucially, if the object is
missing from the second conjunct, this is necessarily in tandem with the finite verb. Furthermore, the construction is
acceptable (more easily so, even) without subject-verb inversion – hence, if no phrase is preposed. For these reasons, it
seems quite plausible that (ii) involves forward ellipsis of the gapping/stripping type. Unfortunately, this conclusion
diverges from the one drawn in the chapter on Asymmetric Coordination, where similar object gaps in German are
discussed.
(ii) %
Vreemd genoeg wilde PEter het huis SLOpen en SuSAN _ renoVEren.
strangely wanted Peter the house demolish and Susan renovate
‘Strangely, Peter wanted to demolish the house and Susan wanted to renovate it.’
(or: ‘... and Peter wanted to renovate Susan’ – not intended) 13
There is another issue related to SLF, namely the possibility of preposing a phrase, say an object, from the first
conjunct only. This seems acceptable in German, but it is questionable in Dutch. It contrasts with the situation in (41),
where the adjunct helaas ‘unfortunately’ can be related to both conjuncts. Clearly, such cases violate the CSC; see
Heycock & Kroch (1994) and Reich (2007) for discussion.
12
(43) a. Peter burned the book [without reading *(it)].
b. Which book did Peter burn _ [without reading (it)]?
Both dependent elements in (42) are c-commanded by the wh-phrase, but there is no c-command
between them. Thus, there appears to be a complex dependency involved, and hence an ATB
configuration in the general sense.14
Similarly, there are parenthetical-like insertions containing an additional gap (cf. Huybregts &
Van Riemsdijk 1985 and De Vries 2009/2013 for some related examples and discussion):
(44) Which proposal did Peter, although in his heart against _, vote in favor of _ ?
Again, there is no regular coordination involved, and the two dependent positions are not linked via
c-command.
This differs from the situation in long-distance movement constructions, in which phrases
may be displaced via intermediate positions in a successive cyclic fashion:
(45) What do you think [ _ (that) Peter bought _ ] ?
Here, the intermediate trace/copy/position does c-command the lower one, and each successive
movement step can be licensed independently. Thus, (45) involves a concatenation of dependencies
rather than a multiple dependency; that is, it is not a one-to-many configuration in the ATB sense.
2.3 Theoretical approaches
Various approaches to ATB can be distinguished, in particular with respect to wh-constructions.
This section contains an overview of the main ideas put forward in the literature.
The central question is how one visible operator can relate to two or more variables. A logical
possibility, depicted in (46), is that all wh-phrases involved are moved separately from their
respective conjuncts, after which each but the first is obligatorily deleted.
(46) Which book (which book) does Peter like _ and Susan hate _ ?
↑ ↑__________________|______________|
|____________________________|
This scenario is highly unlikely, however, because it would involve processes that are otherwise not
available (multiple wh-movement in English, haplology of entire phrases, referential linking of
separate chains), and each individual movement would violate the CSC. Therefore, it appears that
the individual movements need to be collapsed in some way. The picture in (47) represents a
pretheoretical ‘forking movement’ view on ATB.
(47) Which book does Peter like _ and Susan hate _ ?
↑_________________|______________|
Williams’ (1978) solution is dependent on a particular analysis of coordination involving
‘factorization’. In informal terms, this means that the comparable parts of the conjoined clauses are
combined in groups, roughly as illustrated in (48):
14
However, a sideward movement theory of parasitic gaps (cf. Nunes 2001) gives an interesting twist to this
perception. See also the next section.
13
(48) ... | Peter like | which book | and
| Susan hate | which book |
Wh-movement will then place a single copy of the wh-phrase in a higher position (COMP), and
empty the entire factor where it originates.
In this line of thinking, coordination requires some sort of three-dimensional representation.
For a variety of proposals along those lines, but based on different formalizations, see Goodall
(1987), Chametzky (1987), Mu’adz (1991), De Vries (1992), Moltmann (1992), Grootveld (1994),
and De Vries (2005). An important concept used in tandem is structure sharing (which is still
applied in current theories of multidominance; see below). In the simplified representation in (49),
the wh-phrase is shared between both conjuncts, and subsequently moved.
(49) Peter like
Which book does –
Susan hate
The problem of a double operator-variable dependency in ATB constructions vanishes on this view,
since there is only one moved phrase – shared between all conjuncts – and one trace/gap. This is a
major advantage.
In a Minimalist framework, structure sharing can be derived if external remerge (or ‘parallel
merge’) is allowed. Proposals to analyze ATB in such terms include Citko (2005), Kasai (2007),
Gracanin-Yuksek (2007), and De Vries (2013). The resulting representation for the current example
is given in (50), again abstracting away from the details of regular sentence analysis. Here, which
book is merged with each verb successively, and later in the derivation with the Cˊ-projection.
Crucially, such theories are not theoretically multidimensional in the real sense of Williams (1978)
or Grootveld (1994), despite popular belief.
(50)
It must be noted that analyses involving ‘sideward’ movement assume the same derivation
(sequence of mergers), but represent movement in terms of copies instead of multidominance; see
Nunes (2001), Hornstein & Nunes (2002). According to De Vries (2009), they are formally
equivalent.
Approaches involving structure sharing or sideward movement are not generally accepted,
however, for they require a less restrictive grammar and a more complex linearization procedure
than standardly assumed (see, for instance, Van Oirsouw’s 1987 critique of Goodall’s early work).
Furthermore, possible asymmetries in syntactic function and thematic role (cf. section 2.2.2) are
CP
does CoP
Peter
like
and
Susan
hate
which book
14
potentially problematic, since these apparently require the licensing of different sets of features in
one (shared/remerged) phrase. However, both Citko (2011) and De Vries (2012) suggest the
possibility of sharing below the highest DP level, which may be a solution to such issues (i.e., what
is similar in both conjuncts is the shared core; what is dissimilar involves higher functional
projections that are not shared). Furthermore, post-syntax morphology may resort to rescuing
strategies in cases of conflicting feature bundles, causing proximity effects; see Kluck (2009).15
Suggesting a different route, but unfortunately ignoring most of the relevant extraction data,
Sjoblom (1980:27ff) elaborates upon a remark in Chomsky & Lasnik (1977: 491): “On general
grounds, it would be well to explore the possibility that there is no dual extraction from conjoined
clauses; rather, the wh-word that appears derives from the first clause while some sort of deletion
applies in the second.” See also George (1980). But it seems that it is only in Wilder (1994) that the
idea of deletion is taken to its full consequence (see also Salzmann 2012b). Wilder argues that
conjuncts can only be root CPs or DPs, i.e., there are no ‘small conjuncts’ of the type assumed
hitherto. ATB is a ‘byproduct’ of ellipsis (more precisely, left-peripheral deletion) in non-initial
conjuncts. A representation of the simple example used above would be (51a). A slightly more
complex one is in (51b):
(51) a. [Which book does Peter like _ ] and [which book does Susan hate _ ] ?
b. [Who did you say [ that Peter likes _ ]] and [who did you say [ that Susan hates _ ]] ?
Since there is wh-movement within each conjunct separately, ATB as such does not exist within this
approach.
Wilder notes and discusses the pertinent problem of interpretation – which goes back to Ross
(1967) – posed by rejecting ATB and, more generally, small conjuncts (resulting in an abundance of
ellipsis). The point is that a second wh-phrase would normally introduce an additional referent.16
Thus, there must be a referential link between the two instances of who in (51b), for instance.
Wilder proposes that ‘ellipsis chains’ are created, e.g. {who, who}, similar to movement chains.
Here, there is no c-command between the chain members themselves, but Wilder suggests to shift
the attention to the roots (conjuncts) containing them.
Another way of removing the ATB stipulation from the grammar, as well as the CSC
altogether, is pursued in GPSG. In this framework, gaps are indicated by ‘slash categories’. As
Gazdar (1981) points out, the CSC and ATB then simply follow from the like-categories property of
coordination. For instance, a VP from which an NP is displaced is a VP/NP, and as such cannot be
coordinated with a complete VP. If extraction takes place from both VPs, coordination is fine. A
minimal pair from Gazdar (1981:173) involving a relative clause is cited in (52):
(52) a. * The man who Mary loves _ and Sally hates George computed my tax. (= S/NP & S)
b. The man who Mary loves _ and Sally hates _ computed my tax. (= S/NP & S/NP)
Thus, ATB is not viewed as an exception to a stipulated constraint, rather, it represents a normal
instance of symmetry, which naturally facilitates coordination, whereas a breach of (categorial)
symmetry bleeds coordination. Further developments in nontransformational frameworks such as
HPSG and (Combinatory) Categorial Grammar can be found in Gazdar et al. (1985), Steedman
(1985), Bouma, Malouf & Sag (2001), and Chaves (2012), among others.
The simplicity of the argument is quite appealing. Unfortunately, there are also problems
involved, setting aside the issues concerning extraction asymmetries mentioned in section 2.2.5, the
island data in 2.2.3, the non-identity readings discussed in 2.2.4, and the proliferation of categorial
15
See also section 3.3 and the related footnote 19. 16
More generally, interpretation suggests that coordination is always at the lowest possible level (with perhaps
heavily marked exceptions as illustrated in footnote 12). For instance, no books or journals means no [books or
journals], not [no books] or [no journals].
15
labels (including multiple-slashed categories). Pesetsky (1982:554ff) objects that referential indices
on the slashes would be necessary to distinguish them. A relevant example is (53a), which could be
ruled out if coordination is sensitive to referential indices on categorial labels.
(53) a. * a booki that I know whoj to [talk to _j ] and [ buy _i ] (VP/NPi & VP/NPj)
b. John read [a book]i and [a novel]j (NPi & NPj)
But indices on categorial labels seem anomalous; moreover, it is unclear why such feature
assignments would apply to slash categories but not to regular categories, given that (53b) must not
be excluded.
A different approach to ATB that must be mentioned here assumes extraction from the first
conjunct only, and operator movement inside the second, as depicted in (54). See Munn
(1992/1993), and Franks (1993), among others. The coordination phrase contains just the second
conjunct, and is viewed as an adjunct to the first. On this view, the CSC cannot be a syntactic
constraint.
(54) Which book does [Peter like _ ] [OP and Susan hate _ ]?
↑___________________| ↑______________|
Zhang (2010) accumulates evidence against the CSC, but also argues against the adjunction view of
coordination and in favor of a regular coordination phrase (where Co is the head and the conjuncts
are specifier and complement). Nevertheless, her theory of ATB as such shares important features
with Munn’s. In her view, there is a silent pronominal variable inside the second conjunct, which is
bound by the phrase extracted from the first conjunct only; see (55), slightly adapted from Zhang
(2010: 223).17
(55) Which ø<same> book [ _ does Peter like _ ] and [pro-φP does Susan hate _ ]?
↑____________|↑_ _ _ _ _ _ _ _ _| ↑_ _ _ _ _ _ _ _ _ _ _ _|
Zhang stresses the identity reading of such ATB constructions (however, recall from section 2.2.4
that there are also non-identity readings), and explicitly equals the analysis with that of ‘thematic
licensing same constructions’ in coordinate clauses (e.g., The same man got drunk and was arrested
by the cops), presenting this generalization as a major advantage. The example in (56) involves
object topicalization:
(56) The same man [ Mary helped _ ] and [Jane ruined pro-φP].
↑__________________|
However, it remains unclear why anyone would not regard such same-constructions as normal
instances of ATB extraction (see also the overview in section 2.1), and why other analyses could
not cover such cases. Surely, it is true that (57) is not semantically equivalent to (56):
(57) Mary helped the same man and Jane ruined the same man.
But a pre-transormational deep structure is no longer part of generative theories; meaning is
established dynamically or beyond the LF interface. Furthermore, LF-reconstruction after
A´-movement may involve the lexical content of phrases, but not their referential properties. Thus,
17
The additional movements within the conjuncts (indicated by dotted arrows here), due to wh features, are
irrelevant to the general idea. Also note that there is forward deletion of does. Zhang does not provide further details of
clause structure.
16
the identity reading cannot be regarded a problem for movement/multidominance theories of ATB.
See also the discussion below (51).
Finally, it should be mentioned that Salzmann (2012b) combines the idea of asymmetric
extraction from the first conjunct with ellipsis. Like Zhang, he assumes a coordinate structure below
the clause level. Within each conjunct, there is movement of the ATB phrase (of which there are
two instantiations), but it is only extracted from the first. Ellipsis licensing into the second conjunct
is formulated in terms of a specialized [E]-feature, in the spirit of Merchant (2001).
Interestingly, it seems that neither the assumption of asymmetric extraction nor that of sideward
movement/multidominance in ATB hinge on explicit coordinate structures per se. Therefore, the
same mechanisms can potentially be used to explain parasitic gap configurations. Recall from
section 2.2.8 that both can be subsumed under the general definition of ATB as many-to-one
correspondences.
Earlier, Williams (1990) suggested to reduce parasitic gap constructions to ATB movement,
extending the concept of coordination. Postal (1993) shows that this is highly problematic because
ATB is much more general than parasitic gaps are in English. From a cross-linguistic perspective,
ATB movement appears to be wide-spread, but parasitic gaps of the English type are certainly not
universal. For some discussion, see Huybregts & Van Riemsdijk (1985), Rögnvaldsson (1993),
Reich (2007), Niinuma (2010), and Salzmann (2012b), among others. The contrast in (58) is cited
from Postal (1993: 736).
(58) a. How sick did John look _ and (Betty) say he actually felt _ ?
b. * How sick did John look _ without actually feeling _ ?
It is worth noting that there is no principled categorial restriction on selected parasitic gaps,
however; see (59), from Levine, Hukari & Calcagno (2001):
(59) I wonder just how nasty you can pretend to be _ without ACTUALLY becoming _.
While it seems indeed far-fetched to stretch coordination to parasitic gap configurations (or for that
matter ‘reduce’ the mechanism of parasitic gaps to ATB movement or vice versa in some way), a
non-construction-specific theoretical generalization over regular ATB and parasitic gaps does not
logically imply that they must behave similar in every respect. It may very well be that they involve
the same mechanism, but to what extent that mechanism is constrained depends on independent
properties of the particular syntactic environment in which it is applied (e.g., conjuncts versus
adjuncts). See also Hornstein & Nunes (2002) along these lines.
3 Other potential ATB phenomena
While ATB wh-movement is nowadays perceived as the canonical case, it was recognized right
from its inception that ATB patterns are more widespread. This section extends the empirical
domain, taking a bird’s eye perspective.
3.1 Ellipsis
Next to regular instances of movement, certain cases of ellipsis might be considered as ATB,
depending on the particular analyses and definitions.
17
3.1.1 Doing it backwards: Right Node Raising
The introduction already mentioned backward conjunction reduction, which was assumed to
involve rightward ATB movement in Ross (1967) and accordingly dubbed Right Node Raising
(RNR) in Postal (1974: 125). Example (60a) is from Ross (1967: 107), another is (60b), where the
entire verb phrase is related to both conjuncts; (60c) mimics the parenthetical construction in (44):
(60) a. Sally might be, and everyone believes Sheila definitely is, scintillating.
b. Peter may perhaps, and Susan will definitely, buy this book.
c. Peter, although in his heart against, actually voted in favor of, this silly proposal.
On a backward deletion account, a multidimensional representation, or a multidominance analysis
(see also the separate Companion to Syntax chapter on Right Node Raising), the relevant phrase
does not leave the coordination phrase. It is then less clear that such examples represent ATB
because there is no one-to-many dependency (although many-to-one selection might be considered
as inverse ATB; cf. section 3.3).
Differently from ATB movement (cf. section 2.2.3), Right Node Raising is not island
sensitive, see (61), for instance, where the relevant selecting verbs are embedded inside a relative
clause. See De Vries (2013) for recent discussion and further references.
(61) John knows a man who likes, and Peter knows a woman who hates, the prime minister.
However, a clear similarity between ATB movement and ellipsis is that multiplying the process is
straightforwardly possible, but interruptions of parallelism are disallowed; see (62a/b):
(62) a. Peter likes, Susan admires, Mike hates, and John detests, the prime minister.
b. * Peter likes, Susan admires the president, Mike hates, and John detests, the prime
minister.
This can be compared to (31) and (32).
An interesting issue is referential identity. In (63), the referent of a book by Dickens is
normally taken to be constant across conjuncts:
(63) Peter likes, and Susan hates, a book by Dickens.
But a sloppy identity is easily possible in the following examples (see also Ha 2008 and Barros &
Vicente 2011 for discussion):
(64) a. Yesterday, the poor man hired [x], and the rich man bought [y], a red carx resp. y.
b. The beggarj disadvantaged, but the billionairei benefitted, himselfi resp. j.
c. Peteri bumped, and Mikej bruised, hisi resp. j head.
d. After the tennis match, the winning playeri did, but the loserj did not, call hisi resp. j
mother.
Potential speaker and language differences aside, note that these are inherently different from
instances of non-reconstructable (non-distributive, cumulative) RNR; see (65a/b(i)), which are
accepted by some speakers, but clearly rejected by others:
(65) a. (i) %
Peter yelled, and Susan whispered, a different curse.
(ii) Peter yelled a (*different≠ j) cursei and Susan whispered a (*different≠ i) cursej.
b. (i) %
Peter said that Marie, and Mike heard that Susan, have received an award.
(ii) ... Marie has/*have received an award...
18
One might argue that a different structure underlies such cases. See also the phenomenon of split
antecedents in section 3.3.
3.1.2 Forward ellipsis
Williams (1978: 38ff) proposes a specialized ATB mechanism for instances of conjunction
reduction, in particular comparative deletion and gapping; his examples are (66a/b):
(66) a. John has more cows than Bill has dogs or Pete has horses.
b. John gave the book to Mary and the record to Sue, or the book to Sue and the record to
Mary.
In a different fashion, Goodall (1987) and others use structure sharing to account for forward
ellipsis; compare (49) above. What is crucially dissimilar from ATB extraction, is that the
antecedent/licensor is now inside the first conjunct, rather than in a c-commanding position outside
the entire coordinate structure.
A different take on gapping – but not other instances of forward ellipsis – is proposed in
Johnson (2009, and earlier manuscripts), where ellipsis is reduced to ATB movement. The basic
idea is illustrated in (67):
(67) Peters bought [ ts _ a book] and [Susan _ a magazine].
Crucially, the displaced verb (or emptied verb phrase) bought is related to a base position in both
conjuncts. For this to go through, there must be non-parallel movement of the subject from the first
conjunct only, which is not easily justified. See also Kim (2003) and Toosarvandani (2013) for
critique.
3.2 Pronouns, features, and case
One-to-many licensing need not involve extraction (or ellipsis). Binding relationships can be
distributed over conjuncts as well. The examples in (68) illustrate this for reflexive binding and
variable binding, or a combination thereof (for instance, Czech would use reflexive possessive
pronouns in the equivalent of (68b)).
(68) a. Peteri admired [a drawing of himselfi ] and [a painting of himselfi ].
b. Every boyi invited [hisi mother] and [hisi father].
c. No CEOi wondered [whether hei was really competent for the job] and [why hei earned
so much money].
These are all clear ATB configurations. However, unlike the situation with standard ATB
movement, licensing in such cases is not necesssarily across-the-board: the variable may figure only
in a random subset of conjuncts. Compare (69) to (68b), for instance:
(69) a. Every boyi invited [hisi best friend] and [the gym teacher].
b. Every boyi invited [the gym teacher] and [hisi best friend].
Coordination is no prerequisite either. Moreover, the hierarchical configuration in (70) is quite
different from parasitic gap constructions.
(70) Every boyi complained to hisi mother that hisi friends called himi names.
19
Koster (1987: 28) and Neeleman & Van de Koot (2002:548) provide examples involving anaphors
and negative polarity polarity items:
(71) a. They talked with each other about each other.
b. They introduced each other’s financial advisors to each other’s lawyers.
c. John didn’t introduce anyone’s financial advisor to anyone’s lawyer.
Thus, it seems difficult to generalize over ATB extraction and multiple binding.
The next candidate for an ATB status is again quite different: it involves case marking. In
(72a), from German, the preposition zwischen assigns (morphologically visible) dative case to both
coordinated DPs. In English, the same pattern can be shown with accusative pronouns (72b):
(72) a. Er stand zwischen [de-n Tote-n] und [de-n Lebende-n].
he stood between the-DAT dead-DAT and the-DAT living-DAT
‘He stood between the living and the dead.’
b. Peter was standing between him and her.
In (73), the issue is even clearer. In exceptional case marking configurations and for-infinitives, the
relevant case assignment involves a subconstituent of both conjuncts:
(73) a. Peter saw [him coming] and [her going].
b. Peter believes [him to be nice] and [her to be bad].
c. What Peter wants is for [him to leave] and [her to stay].
While some of the examples might be explained away by invoking forward conjunction reduction
of the left edge (e.g., ... and believes her to be bad in (73b)),18
this is certainly not a general
solution. First, plural-taking predicates cannot be treated in this way, cf. we compared him and
(*compared) her; or between him and (*between) her. Second, the coordination as a whole can
sometimes be moved away or isolated from the selecting predicate: for instance, [Him coming and
her going] is what Peter saw.
Despite that, one might object that the relevant examples are not convincing instances of one-
to-many relationships because the dependency may be between the case assigner and the entire
coordinated phrase. However, this is slightly besides the point. In a similar vein, standard ATB wh-
movement as in (1) can be explained as a relationship between the wh-constituent and a complex
category containing a combined slash/gap feature. What is relevant is that a coordination phrase is
apparently able to distribute properties over its conjuncts (from a top-down perspective), or
accumulate properties from its members (from a bottom-up perspective).
3.3 Split antecedents and other ‘inverse ATB’ phenomena
Finally, it may be worth considering whether there is ‘inverse ATB’ in the sense of many-to-one
dependencies. In this respect, agreement with coordinated noun phrases could be relevant. In (74),
the finite verb shows number agreement with the conjoined subject. In French, and many other
languages, adjectives show phi-feature agreement as well; see (75):
(74) John and Jack are busy.
(75) Marie et Claire sont dangereus-e-s.
Marie and Claire are dangerous-FEM-PL
18
Left edge deletion implies coordination at a higher level, which seems to be at variance with general economy
considerations (‘simplest structure’); see also footnote 16.
20
Notice, however, that the plural marking in these examples cannot be traced back to the separate
conjuncts: it is not a distributed property, but an emergent property of the coordination as a whole.
This is different from standard ATB, where the distribution of a grammatical relationship over the
separate conjuncts is essential. In (75), however, the feature [feminine] can indeed be related to both
conjuncts individually. It then needs to be proven that the relationship is not in fact an instance of
first or second conjunct agreement. In this particular case, that seems to be true, for if either the first
or second conjunct represents a male, the adjective or participle would get a default masculine
gender inflection, which is zero. Interestingly, the relevant feminine agreement is retained on a
participle in French even if the coordinated subject is inherently distributive, as can be shown by
means of an initial coordinator. Example (76) is cited and translated from a random internet blog.
(76) Ni les frontière-s de 1948, ni celle-s de 1967 n’ont eté
neither the borders:FEM-PL of 1948 nor those:FEM-PL of 1967 NEG-have been
reconnu-e-s comme limites à l’expansion israélienne.
acknowledged-FEM-PL as limits to the-expansion Israeli
‘Neither the 1949 borders nor those of 1967 have been acknowledged as limits to the Israeli
expansion.’
Generally, agreement resolution across languages and construction types is a complicated matter,
and hasty conclusions should be avoided.19
An abstractly related state of affairs is attested in cases with split ancedents. In (77), the
dependent element, the relative pronoun who, emerges from the joint interpretation of the separated
heads/antecedents (equivalently to the plural pronoun they).20
Whether the predicate in the relative
clause is collective or inherently individual is irrelevant, as is made explicit in (77b).
(77) a. A boyi arrived and a girlj left whoi+j hated each other.
b. I saw a mani in the house and I saw a womanj in the garden whoi+j were {married to
each other, (each/both) chewing gum}.
Although the relative clause must be peripheral, such examples are not to be confused with true
cases of RNR: in (78), the singular verb in the relative clause and the particular contrastive
intonation pattern are only compatible with a distributive reading, here enforced by the use of the
not only...but also combination:21
(78) I not only saw a mani in the HOUSE, but also a womanj in the GARDEN, whoi/j was
{chewing gum, *married to each other, *each chewing gum}.
19
See Corbett (2006) for an accessible overview. Remarkably, even in a language with a relatively poor morphology
such as Dutch, there can be a whole variety of agreement issues related to coordination or feature doubling otherwise.
For elaborate discussion, see e.g. Heringa & De Vries (2008), De Vries & Heringa (2008), Kluck (2009), De Vries
(2004); dialectal variation and complementizer agreement is examined in Van Koppen (2005, 2012). 20
It is far from obvious how such constructions are to be derived, as was first noted in a one-page squib by
Perlmutter & Ross (1970). See Hoeksema (1986) for a more elaborate discussion, and Zhang (2007, 2010) and
McKinney-Bock (2013) for more recent proposals and further references. Notice also that a comparable problem shows
up even in seemingly simple cases as (i), called hydrae by Link (1984). Since the relative clause semantically intersects
with the noun (phrase) below the level of the determiner, the expected configuration is (ii) – also acceptable –, where
nominal properties can be conjoined at the right level, as indicated with subscripts. In (i), the intervening determiner
creates the problematic split.
(i) the boyi and the girlj whoi+j met yesterday
(ii) the [[boyi and girlj]k whok=i+j met yesterday] 21
See Kluck & De Vries (2013) for related data from Dutch, and further discussion. In the type of example in (78),
RNR is preceded by extraposition.
21
Notice that on a more neutral intonation, there is a salient alternative reading in which the relative
clause is related to a woman only; in that case there is no split antecedent at all. Other factors to
consider, likely also reflected in the intonation, are the mode of construal of both the relative clause
and the second conjunct. The former can alternatively be interpreted appositively, and the conjoined
clause can be construed as a parenthetical. A study in which all these complications are carefully
sorted out is still due.
To complete the puzzle, it is also possible to create a crossing one-to-one dependency for
some speakers (cf. Dougherty 1969), as in (79). Note that such data are considered unacceptable by
many others (see also George 1980: 120ff for discussion).
(79) %
Peter admires, and Mike worships, Susan and Anne, respectively.
intended: ‘Peter admires Susan, and Mike worships Anne.’
Superficially, the examples in (80a/b) involve a similar configuration. However, the simple example
in (80c) suggests that the issue of matching the separate actions with the right individuals is
pragmatic rather than syntactic or semantic in these cases:
(80) a. %
I saw a man in the house and I saw a woman in the garden who were chewing gum and
drinking coke, respectively.
b. A man and a woman were chewing gum and drinking coke, respectively.
c. They were chewing gum and drinking coke, respectively.
Here, who, a man and a woman, and they all represent a plural subject. By contrast, Peter and Mike
in (79) are not combined into a plural, so a complex form of RNR is called for.22
In case of a split antecedent as in (77), the two antecedents are not directly coordinated to
each other; still, there is coordination at a higher level. This raises the question of whether this is
required. The following examples suggest that this is not the case. In (81a), the silent PRO can take
split antecedents that are in a c-command configuration for many speakers, very much comparable
to what a personal pronoun can do (81b), depending on binding conditions; similarly, the relative
pronoun in (81c) can relate to a combination of noun phrases:
(81) a. %
Peteri proposed to Susanj to PROi+j go to the cinema together.
b. Peteri said to Susanj that theyi+j could go to the cinema together.
c. %
A mani kissed a womani yesterday whoi+j actually hated each other.
The simple generalization seems to be that every pro-form can pick up any salient information of
the right type as its antecedent, split or not, and that this has nothing to do with syntactic
coordination per se. The additional difficulty in (81c) and (77/78) is that the relative clause acts as a
restriction on the interpretation of (each of) the nominal antecedents.
The issue of split dependencies is even broader than indicated so far. In (82) and (83), the
result clauses and comparative clauses depend on split degree terms.23
See also Chomsky
(1981:81ff) for some discussion.
(82) a. Peter is so tall and Susan is so small [that they can hardly look each other straight in the
eye].
b. So many thieves stole so many diamonds each [that the insurance could not possibly
cover the loss].
22
See Park (2005), De Vos & Vicente (2005), and Zhang (2010) for related proposals. 23
Evidently, such examples are problematic if the clause is to be generated within the maximal projection of the
degree element. See Rijkhoek (1998) for an alternative approach.
22
(83) a. More boys wanted to sing and more girls wanted to dance [than there were places in the
joint competition].
b. More students asked more questions about more topics to more professors on the panel
[than the organizers had imagined in their wildest dreams].
As before, it does not seem to matter whether these terms are embedded in coordinated clauses or
not.
4. Conclusion
Across-the-board phenomena in the broad sense of one-to-many and many-to-one relationships are
surprisingly widespread throughout the language system. Whether or not a particular construction
type is to be characterized as ‘true’ ATB ultimately depends on the precise analysis of the
construction as well as the definition and theory of ATB itself. While there is a common trait to
ATB and ATB-like constructions, it seems improbable that a generalized analysis can cover all of
them. A number of different classes may be distinguished. First, there is ATB extraction from
coordinate structures. These encompass any kind of leftward movement simultaneously from more
than one conjunct. A, A´ and head movement are all attested, and there is no obvious limitation with
respect to categorial status. Second, multiple extraction from subordination contexts is possible;
these include parasitic gap configurations. Third, ellipsis phenomena, forward and backward in
different ways, share some properties with ATB extractions. Fourth, binding of pronouns may
involve multiple dependencies. Fifth, in inverse fashion, this may result in split antecedent
phenomena. Sixth, Case and agreement patterns often target multiple conjuncts.
ATB was originally presented as an exception to the coordinate structure constraint. There
seems to be a growing consensus that the CSC cannot be part of syntax, and hence that coordinate
structures are not inherent syntactic islands. ATB as well as extractions from semantically
asymmetric coordinations are then no longer exceptions to a constraint. The remaining effect of the
CSC comes down to a certain parallelism requirement concerning ‘symmetrical’ coordination,
which may be a semantic effect. In these structures, and only there, extraction must take place from
all conjuncts. Similarly, forward and backward ellipsis necessarily targets all conjuncts save the
first or last where the antecedent resides. However, no such demand exists for binding of pronouns
inside a coordination phrase. Other potential parallelism constraints have been discussed in the
literature, but none are really robust. A related issue is the so-called (referential) identity
requirement in ATB constructions, which was shown to be invalid here.
It appears to be a given that the grammar allows for multiple dependencies, as is especially
clear from multiple binding inside and outside coordination contexts. But this alone is insufficient
to explain ATB extraction in a transformational framework. There are basically two approaches.
The first states that only the movement from the first base position is real, the other gaps involve a
coreferent operator or silent pronoun. While this is a feasible way to go, it is somewhat
discomforting that there is little convincing evidence for such an asymmetrical configuration;
moreover, the empty elements need to be justified. The alternative involves structure sharing and/or
sideward movement – if one accepts that such theoretical options are independently available. Such
analyses directly reflect the intuitive symmetry of the configuration, but they face some problems of
execution as well. Notably, both approaches can easily generalize over regular ATB in coordinate
structures and parasitic gaps. The old question of whether parasitic gaps are to be reduced to ATB
or vice versa is an inadequate way of framing the issue. It is likely that a common mechanism
underlies (apparent) multiple extraction – but it is still necessary to explain the differences in
behavior (to the extent they are real), plausibly in terms of the different configuration in which the
mechanism is applied.
23
Evidently, a full evaluation of all existing theoretical approaches in the light of all available
data is impossible within the limits of this chapter, and might even be considered unjustified from a
historical perspective. Nevertheless, for future progress in the study of ATB phenomena it may be
fruitful to continue theoretical developments with an eye on the expanded empirical basis along the
lines indicated here.
Cross-References
See also Asymmetric Coordination, Conjoined Wh-Questions, Parasitic Gaps, Right Node Raising
References
Barros, Matt & Luis Vicente. 2011. Right Node Raising Requires both Ellipsis and
Multidomination. University of Pennsylvania Working Papers in Luinguistics 17.
Bošković, Željko & Steven Franks. 2000. Across-the-board movement and LF. Syntax 3(2),107-
128.
Bouma, Gosse, Rob Malouf & Ivan Sag. 2001. Satisfying constraints on extraction and adjunction.
Natural Language & Linguistic Theory 19, 1-65.
Chametzky, Robert. 1987. Coordination and the Organization of a Grammar. PhD Dissertation,
University of Chicago.
Chaves, Rui. 2009. Construction-based Cumulation and Adjunct Extraction. Proceedings of the
16th International Conference on Head-Driven Phrase Structure Grammar, 47-67. Stanford,
CA: CSLI Publications.
Chaves, Rui. 2012. On the grammar of extraction and coordination. Natural Language &
Linguistic Theory 30, 465-512.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam & Howard Lasnik. 1977. Filters and Control. Linguistic Inquiry 8, 425-504.
Citko, Barbara. 2005. On the nature of Merge: External Merge, Internal and Parallel Merge.
Linguistic Inquiry 36, 475-497.
Citko, Barbara. 2011. Symmetry in Syntax: Merge, Move, and Labels. Cambridge University Press.
Corbett, Greville. 2006. Agreement. Cambridge University Press.
Culicover, Peter & Ray Jackendoff. 1997. Semantic subordination despite syntactic coordination.
Linguistic Inquiry 28, 195-217.
De Vos, Mark. 2005. The syntax of pseudo-coordination in English and Afrikaans. PhD
Dissertation, Leiden University. LOT Dissertation Series 114.
De Vos, Mark & Luis Vicente. 2005. Coordination under Right Node Raising. Proceedings of the
24th West Coast Conference on Formal Linguistics, eds. John Alderete et al., 97-104.
Somerville, MA: Cascadilla.
Dotlačil, Jakub. 2008. Across-the-Board Extraction and Clitic Omission in Czech. Proceedings of
Formal Approaches to Slavic Linguistics 16: The Stony Brook Meeting, eds. A. Antonenko, J.
Bailyn & Bethin. Ann Harbor: Michigan Slavic Publications, 134-150.
Dougherty, Ray. 1969. A transformational grammar of coordinate conjoined structures. PhD
Dissertation, MIT.
Dyła, Stefan. 1984. Across-the-board dependencies and case in Polish. Linguistic Inquiry 15, 701-705.
Engdahl, Elisabet. 1983. Parasitic gaps. Linguistics and Philosophy 6, 5-34.
Franks, Steven. 1993. On parallelism in across-the-board dependencies. Linguistic Inquiry 24, 509-
529.
Gawron, Jean Mark & Andrew Kehler. 2004. The semantics of respective readings, conjunction, and
filler-gap dependencies. Linguistics and Philosophy 27, 169-207.
Gazdar, Gerald. 1981. Unbounded dependencies and coordinate structure. Linguistic Inquiry 12, 155-
183.
24
Gazdar, Gerald et al. 1985. Generalized Phrase Structure Grammar. Cambridge, MA: Harvard
University Press.
George, Leland. 1980. Analogical generalization in natural language syntax. PhD Dissertation, MIT.
Goldsmith, John. 1985. A principled exception to the Coordinate Structure Constraint. Proceedings of
CLS 21, Part 1, 133-143. Chicago Linguistic Society, University of Chicago, Ill.
Goodall, Grant. 1987. Parallel structures in syntax: Coordination, causatives and restructuring.
Cambridge University Press.
Gracanin-Yuksek, Martina. 2007. About Sharing. PhD Dissertation, MIT.
Grootveld, Marjan. 1994. Parsing Coordination Generatively. PhD Dissertation, Leiden University.
Grosu, Alexander. 1973. On the nonunitary nature of the Coordinate Structure Constraint. Linguistic
Inquiry 4, 88-92.
Ha, Seungwan. 2008. Ellipsis, Right Node Raising, and Across-the-Board constructions. PhD
Dissertation, Boston University.
Haïk, Isabelle. 1985. The Syntax of Operators. PhD Dissertation, MIT.
Heringa, Herman & Mark de Vries. 2008. Enkelvoudige congruentie bij nevenschikking met en:
puzzels met betrekking tot massa’s en demonstratieven [Single agreement involving and-
coordination: puzzles with respect to mass nouns and demonstratives]. Tabu 37, 19-38.
Heycock, Caroline & Anthony Kroch. 1994. Verb movement and coordination in a dynamic theory of
licensing. The Linguistic Review 11, 257-283.
Heycock, Caroline & Roberto Zamparelli. 2000. Friends and colleagues: plurality and NP-
coordination. Proceedings of NELS 30, 341-352.
Hicks, Glyn. 2009. Tough-constructions and their derivation. Linguistic Inquiry 40, 535-566.
Hoeksema, Jack. 1986. An account of relative clauses with split antecedents. Proceedings of the 5th
West Coast Conference on Formal Linguistics (WCCFL 5), eds. Mary Dalrymple et al., 68-86.
Stanford, CA: CSLI Publications.
Höhle, Tilman. 1983. Subjektslücken in Koordinationen [Subject holes in coordination]. Manuscript,
University of Cologne.
Hornstein, Norbert & Jairo Nunes. 2002. On asymmetries between parasitic gap and Across-The-
Boaord constructions. Syntax 5, 26-54.
Huybregts, Rini & Henk van Riemsdijk. 1985. Parasitic gaps and ATB. Proceedings of NELS 15, eds.
S. Berman, J.-W. Choe & J. McDonough, 168-187.
Johannessen, Janne Bondi. 1998. Coordination. Oxford University Press.
Johnson, Kyle. 2009. Gapping is not (VP-)Ellipsis. Linguistic Inquiry 40, 289-328.
Lakoff, George. 1986. Frame semantic control of the Coordinate Structure Constraint. Proceedings of
CLS 22, part 2, 152-167. Chicago Linguistic Society, University of Chicago, Chicago, Ill.
Kasai, Hironobu. 2004. Two notes on ATB movement. Language and Linguistics 5, 167-188.
Kasai, Hironobu. 2007. Multiple dominance in syntax. PhD Dissertation, Harvard University.
Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Kehler, Andrew. 2002. Coherence, reference, and the theory of grammar. Stanford: CSLI
Publications.
Kim, Jeong-Seok. 2003. A reply to Johnson’s (1994) across-the-board movement analysis of gapping.
Studies in Generative Grammar, Vol. 13, 143-147.
Kluck, Marlies. 2009. Good neighbors or far friends. Matching and proximity effects in Dutch Right
Node Raising. Groninger Arbeiten zur Germanistischen Linguistik 48, 115-158.
Kluck, Marlies & Mark de Vries. 2013. Cumulative rightward processes. Rightward Movement in a
Comparative Perspective, eds. Gert Webelhuth, Manfred Sailer & Heike Walker, 281-317.
Amsterdam: John Benjamins.
Koopman, Hilda & Dominique Sportiche. 1982. Variables and the bijection principle. The Linguistic
Review 2, 139-160.
van Koppen, Marjo. 2005. One Probe - Two Goals: Aspects of agreement in Dutch dialects. PhD
Dissertation, Leiden University. LOT Dissertation Series 105.
25
van Koppen, Marjo. 2012. The distribution of phi-features in pronouns. Natural Language &
Linguistic Theory 30, 135-177.
Koster, Jan. 1987. Domains and dynasties. Dordrecht: Foris.
Koster, Jan. 1999. The Word Orders of English and Dutch: Collective vs. Individual Checking.
Groninger Arbeiten zur germanistischen Linguistik 43, 1-42.
Koster, Jan. 2003. Move, Merge and Percolate are One! On the Elimination of Variables in
Grammar. Grammatik i Fokus. Grammar in Focus: A Festschrift for Christer Platzack, eds.
L.-O. Delsing, C. Falk, G. Josefsson & H. Sigurðsson, Department of Scandinavian Languages,
University of Lund. < http://www.let.rug.nl/koster/papers/CP-Koster3!.pdf >
Levine, Robert 2001. The extraction riddle: just what are we missing? Journal of Linguistics 37,
145-174.
Levine, Robert, Thomas Hukari & Michael Calcagno. 2001. Parasitic gaps in English: Some
overlooked cases and their theoretical implications. Parasitic Gaps, eds. Peter Culicover & Paul
Postal, 181-222. Cambridge, MA: The MIT Press.
Link, Godehard. 1984. Hydras: On the Logic of Relative Constructions with Multiple Heads.
Varieties of Formal Semantics, Proceedings of the fourth Amsterdam Colloquium, September
1982, eds. F. Landman & F. Veltman. Dordrecht: Foris, 245-257.
Matos, Gabriela. 2000. Across-the-Board clitic placement in Romance languages. Probus 12, 229-
259.
McKinney-Bock, Katherine. 2013. Deriving Split-Antecedent Relative Clauses. University of
Pennsylvania Working Papers in Linguistics 19, 113-122.
Merchant, Jason. 2001. The Syntax of Silence: Sluicing, Islands and the Theory of Ellipsis. New
York: Oxford University Press.
Merchant, Jason. 2003. Remarks on stripping. Ms., University of Chicago.
Moltmann, Friederike. 1992. Coordination and comparatives. PhD Dissertation, MIT.
Mu’adz, Husni. 1991. Coordinate structures: A planar representation. PhD Dissertation, University
of Arizona.
Munn, Alan. 1992. A null operator analysis of ATB gaps. The Linguistic Review 9,1-26.
Munn, Alan. 1993. Topics in the syntax and semantics of coordinate structures. PhD Dissertation,
University of Maryland.
Munn, Alan. 1999. On the identity requirement of ATB extraction. Natural Language Semantics 7,
421-425.
Neeleman, Ad & Hans van de Koot. 2002. The Configurational Matrix. Linguistic Inquiry 33, 529-
574.
Niinuma, Fumikazu. 2010. Across-the-Board and Parasitic Gap Constructions in Romanian.
Linguistic Inquiry 41, 161-169.
Nunes, Jairo. 2001. Linearization of Chains and Sideward Movement. Cambridge, MA: MIT Press.
van Oirsouw, Robert. 1987. Three dimensionality. Formal parameters of generative grammar,
Yearbook III, eds. G. de Haan & W. Zonneveld, 31-46. Utrecht University.
Park, Myung-Kwan. 2005. When things are cumulated or distributed across coordinate conjuncts.
Studies in Generative Grammar 15, 415-431.
Perlmutter, David & John Ross. 1970. Relative clauses with split antecedents. Linguistic Inquiry
1:350.
Pesetsky, David. 1982. Paths and categories. PhD Dissertation, MIT.
Rijkhoek, Paulien. 1998. On degree phrases and result clauses. PhD Dissertation, University of
Groningen.
Postal, Paul. 1974. On raising. Cambridge, MA: MIT Press.
Postal, Paul M. 1993. Parasitic gaps and the across-the-board phenomenon. Linguistic Inquiry 24,
734-754.
Postal, Paul. 1998. Three investigations of extraction. Cambridge, MA: MIT Press.
Reich, Ingo. 2007. From phases to “across the board movement”. Proceedings from the Annual
Meeting of the Chicago Linguistic Society, Vol. 43, No. 2, 217-232.
26
Rögnvaldsson, Eiríkur. 1993. Coordination, ATB-extraction, and the identification of pro. Harvard
Working Papers in Linguistics 3, 153-180.
Ross, John. 1967. Constraints on variables in Syntax. PhD Dissertation, MIT. Published as Infinite
syntax! Norwood: Ablex, 1986.
Sabbagh, Joseph. 2007. Ordering and linearizing rightward movement. Natural Language &
Linguistic Theory 25, 349-401.
Salzmann, Martin. 2012a. Deriving reconstruction asymmetries in Across The Board movement by
means of asymmetric extraction + ellipsis. Comparative Germanic Syntax: The state of the art,
eds. P. Ackema et al. Amsterdam: John Benjamins, 353-385.
Salzmann, Martin, 2012b. A derivational ellipsis approach to ATB-movement. The Linguistic
Review 29, 397-438.
Schmerling, Susan. 1972. Apparent counterexamples to the Coordinate Structure Constraint: a
canonical constraint. Studies in the Linguistic Sciences 2, 91-104.
Sjoblom, Todd. 1980. Coordination. PhD Dissertation, MIT.
Steedman, Mark. 1985. Dependency and Coördination in the Grammar of Dutch and English.
Language 61, 523-568.
Taraldsen, Knut. 1981. The Theoretical Interpretation of a Class of Marked Extractions. Theory of
Markedness in Generative Grammar, eds. A. Belletti, L. Brandi & L. Rizzi, 475-516. Pisa:
Scuola Normale Superiore.
Toosarvandani, Maziar. 2013. Gapping is low coordination (plus VP-ellipsis): A reply to Johnson.
Manuscript, MIT.
de Vries, Gertrud. 1992. On Coordination and Ellipsis. PhD Dissertation, Catholic University of
Brabant [Tilburg University].
de Vries, Mark. 2002. The Syntax of Relativization. PhD Dissertation, University of Amsterdam.
LOT Dissertation Series 53.
de Vries, Mark. 2004. Congruentie-effecten in uitbreidende en vrije relatieve zinnen [Agreement
effects in nonrestrictive and free relative clauses]. Nederlandse Taalkunde 9, 29-47.
de Vries, Mark. 2005. Coordination and Syntactic Hierarchy. Studia Linguistica 59, 83-105.
de Vries, Mark. 2009. On Multidominance and Linearization. Biolinguistics 3, 344-403.
de Vries, Mark. 2012. Unconventional Mergers. Ways of Structure Building, eds. Myriam Uribe-
Etxebarria & Vidal Valmala, 143-166. Oxford: Oxford University Press.
de Vries, Mark. 2013. Multidominance and Locality. Lingua 134, 149-169.
de Vries, Mark & Herman Heringa. 2008. Congruentie met nevengeschikte subjecten: de invloed
van distributiviteit [Agreement with conjoined subjects: the influence of distributivity]. Tabu 37,
3-18.
Wilder, Chris. 1994. Coordination, ATB, and ellipsis. Groninger Arbeiten zur germanistischen
Linguistik 37, 291-331.
Williams, Edwin. 1978. Across-the-board rule application. Linguistic Inquiry 9, 31-43.
Williams, Edwin. 1990. The ATB theory of parasitic gaps. The Linguistic Review 6, 265-279.
Woolford, Ellen. 1987. An ECP account of constraints on Across-The-Board extraction. Linguistic
Inquiry 18, 166-171.
Zhang, Niina Ning. 2004. Against Across-The-Board Movement. Concentric: Studies in Linguistics
30, 123-156.
Zhang, Niina Ning. 2007. The syntactic derivations of split antecedent relative clause constructions.
Taiwan Journal of Linguistics 5, 19-48.
Zhang, Niina Ning. 2010. Coordination in Syntax. Cambridge University Press.
Zwart, Jan-Wouter. 1991. Subject deletion in Dutch: a difference between subjects and topics.
Language and Cognition 1 [yearbook], eds. M. Kas, E. Reuland, & C. Vet, 333-350.
Zwart, Jan-Wouter. 2012. Easy to (re)analyse: tough-constructions in minimalism. Linguistics in
The Netherlands 29, 147-158.